TorontoCL at CMCL 2021 Shared Task: RoBERTa with Multi-Stage Fine-Tuning for Eye-Tracking Prediction

TL;DR

This paper presents a multi-stage fine-tuning approach using RoBERTa for predicting eye-tracking features during reading, achieving competitive results in a shared task.

Contribution

The authors introduce a two-stage fine-tuning method with ensembling on Transformer models for eye-tracking prediction.

Findings

Achieved a MAE score of 3.929, ranking 3rd in the shared task.

Demonstrated the effectiveness of multi-stage fine-tuning and ensembling.

Compared different Transformer models for improved performance.

Abstract

Eye movement data during reading is a useful source of information for understanding language comprehension processes. In this paper, we describe our submission to the CMCL 2021 shared task on predicting human reading patterns. Our model uses RoBERTa with a regression layer to predict 5 eye-tracking features. We train the model in two stages: we first fine-tune on the Provo corpus (another eye-tracking dataset), then fine-tune on the task data. We compare different Transformer models and apply ensembling methods to improve the performance. Our final submission achieves a MAE score of 3.929, ranking 3rd place out of 13 teams that participated in this shared task.